Diabetes is a chronic condition whose treatment and complications account for a disproportionate share of health care expenditures. Prior evaluations of direct and indirect costs of diabetes have been limited by incomplete measures of costs, inappropriate attribution of complications to diabetes, or by inappropriate methodology. We propose to provide estimates of the marginal direct and indirect costs of diabetes in a unified econometric framework using data from the 1987 National Medical Expenditure Survey. The econometric model will have a block-recursive structure consisting of three blocks. The first block will model the contribution of diabetes to certain health conditions using a multivariate probit model. The estimates from this model will be used to create predicted probabilities that will be used as controls in the subsequent blocks. The next block will consist of a two-part model of the health care expenditures for outpatient care, hospital care and nursing facility care. This will provide individual-specific estimates of the incremental health care costs due to diabetes. The third block will model three aspects of labor market activity: labor force participation, number of hours worked, and wages earned. This will provide estimates of the marginal contribution of diabetes to diminished labor market activity. Finally, the health care and labor market costs for each individual will be combined to obtain the aggregate marginal cost of diabetes. Estimates of these costs for typical individuals with socioeconomic characteristics that are of concern to policy makers will also be provided. Our results will provide a basis for sound public policy decisions, including the potential benefit of preventive care.